Resistance to chemotherapy occurs in cancer cells because of intrinsic or acquired changes in expression of specific proteins. We have studied resistance to natural product chemotherapeutic agents such as doxorubicin, Vinca alkaloids, and taxol and more recently, histone deacetylase inhibitors and targeted kinase inhibitors. In most cases, cells become simultaneously resistant to multiple drugs because of reductions in intracellular drug concentrations. For the natural product drugs, this cross-resistance is frequently due to expression of an energy-dependent drug efflux system (ABC transporter) known as P-glycoprotein (P-gp), the product of the MDR1 or ABCB1 gene, or to other members of the ABC transporter family, including ABCG2 and ABCB5. Work from our laboratory and others has revealed that some drugs are more toxic to P-gp-expressing cells than to non-expressors, suggesting a novel approach to treatment of MDR cancers. Several different chemical classes with this property, including thiosemicarbazones (e.g., NSC73306), have been identified. A quantitative structure activity analysis of NSC73306 analogs, a further correlation analysis in the NCI-60 cell lines, and a high-throughput screen for compounds in the U.S. Pharmacopeia that kill P-gp-expressing cells have yielded many additional compounds with improved ability to kill selectively P-gp-expressing cells, but also with improved solubility properties. Not only are ABC transporters responsible for drug resistance in cancer, but they are a major component of the blood-brain barrier (BBB) and blood-placental barrier. The three most prominent transporters at the blood-brain barrier are ABCB1, ABCC1, and ABCG2. We have developed a system for analysis of ABCG2 expression at the blood-brain and the blood-placental barriers based on the fact that luciferin is an ABCG2 substrate at these barriers and its passage into the brain or into developing fetuses can be detected in transgenic mice in which luciferase is expressed at the blood-brain barrier or blood-placental barrier. Because studies of the BBB in mice are time-consuming and expensive, we are developing a parallel analysis in zebrafish, as components of the zebrafish BBB appear to be very similar to those of the mammalian BBB. Zebrafish do not have a direct homolog of human ABCB1 but instead have 2 similar variants-Abcb4 and Abcb5. Expression of these transporters in heterologous systems has enabled their detailed characterization and inhibition properties. In collaboration with Matthew Hall at NCATS, we have found that zebrafish Abcb4 is nearly identical to human ABCB1 in conferring resistance to nearly 100 known ABCB1 substrates. Localization of the transporters to the zebrafish BBB is in progress. Zebrafish also have 4 homologs of human ABCG2-Abcg2a, Abcg2b, Abcg2c and Abcg2d. These transporters have been expressed and a detailed characterization of their substrate specificity is underway. Early results suggest zebrafish Abcg2a and Abcg2c have overlapping substrate specificities with human ABCG2. To understand how the structure of P-gp determines its polyspecificity and how specificity is altered with changes in folding, we have collaborated with other senior investigators in the LCB, including Di Xia, Suresh Ambudkar, and Sriram Subramaniam. Cryo-EM studies have demonstrated that apo P-gp has a dynamic structure in which the two ATP-binding sites are either separated or close together. Binding of ATP fixes the conformation of P-gp in the latter state and ATP hydrolysis results in separation of the ATP sites. Crystallography studies using mouse P-gp as a model show that the separation between the ATP sites determines the pitch of the transmembrane (TM) helices where substrates bind, suggesting the hypothesis that as the ATP sites move together and apart, the TM helices expose different residues that enable binding to many different substrates. Studies on mouse-human chimeric P-gps have revealed similar structure-function relationships for these two evolutionarily related transporters. We have created a highly sensitive, quantitative assay for ABC transporter mRNAs and other mRNAs associated with drug resistance in cultured cancer cells. We have studied samples from human ovarian cancer, hepatocellular cancer (HCC), and acute myelogenous leukemia (AML) in some detail. In ovarian cancer, there is an 11-gene MDR signature associated with poor response to chemotherapy. In HCC, the signature is more complex, but accurately distinguishes poor prognosis vs. better prognosis cancer. Agents that change the pattern of gene expression from poor prognosis to better prognosis patterns also sensitize cultured HCC cells to anti-cancer drugs. Many of these agents are histone deacetylase inhibitors. For AML, samples from the same patients before and after chemotherapy were studied. In this case, resistance in each case shows a different pattern of expression of ABC genes and other MDR genes, suggesting that individualized approaches to resistance to therapy will be needed. A more detailed analysis of a large population of primary refractory AMLs has shown that there are 3 molecular signatures that predict poor response to therapy. One of these is associated with increased expression of P-gp. ABCB5 is a close molecular relative of ABCB1. It is expressed in pigmented cells in the brain and eye, and in melanoma. In collaboration with Richard Cannon (University of Otega, New Zealand) we have shown that when expressed in yeast, ABCB5 is a multidrug transporter. ABCB5 knock-out mice are fully viable, but sensitive to the major tranquilizer haloperidol, consistent with a role of this transporter in the brain (with Gary Peltz, Stanford School of Medicine). Mutations in ABCB5 are found in some melanoma patients and appear to be associated with the malignant phenotype.